Semantically mediated priming of phonology is a delicate phenomenon-what O' Seaghdha and Marin (1997) called "a real but slender effect." In semantically mediated phonological priming, a prime word such as night activates the phonology of a semantically related mediating word, day. In turn, this activation affects naming performance to a target like dare, which partially shares the phonologyof day. Thus, the indirectly related prime, night, primes the target, dare. In their study, O'Seaghdha and Marin (1997) found a weak facilitation of naming response from semantically mediated phonological priming: Participants were faster to name targets preceded by an indirectly related prime than those preceded by an unrelated prime. However, this difference was usually only statistically reliable by participants, not by items.O'Seaghdha and Marin (1997) explicitly tied the empirical delicacy of semantically mediated phonologicalpriming to mediating lexical nodes (see also Dell & O'Seaghdha, 1991). In normal associative priming (e.g., night primes day), activationspreads directly from the target to the prime. However, in semantically mediated phonological priming, the spread of activation is indirect: Night activates the mediating word day, which then partly activates the phonology of dare. According to O'Seaghdha and Marin (1997), it is this extra distance "over noisy lines" (p. 250) between lexical nodes that attenuates the effect of semantically mediated phonologicalpriming. In this article, we suggest that semantically mediated phonologicalpriming is not always delicate. Rather, we use the phonologicalcoherence hypothesis, which assumes that the activation of a word's phonology is derived from direct interactive activation between orthographic, phonologic, and semantic nodes, to predict that relatively robust semantically mediated phonological priming can be found if the target words have the appropriate spelling-phonology relationships (i.e., targets with a phonology that coheres relatively slowly).Semantically mediated phonological priming is important because it provides a good test of interactiveactivation models (Dell, 1986(Dell, , 1988MacKay, 1987;Stemberger, 1985). In the cited accounts, which assume the presence of lexical nodes, activation flows in both directions between phonology nodes and lexical nodes and between lexical nodes and semantic features. Consequently, we should find a variety of effects in which activation spreads all the way from semantic features to phonology nodes (for example) or all the way from phonology nodes to semantic features. The latter category of effects is well established. Targets (e.g., look) are named faster when preceded by pseudoprimes that sound like semantic associates (e.g., stair), as compared with control primes (Lesch & Pollatsek, 1993), and pseudohomophone primes work as well as word primes (e.g., tode primes naming of frog; Lukatela & Turvey, 1994). Three experiments test for semantically mediated priming of a word's phonology (e.g., sofa, an associate of couch, pri...